See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265629382

Simulation and Control of Spray Drying Using Nozzle Atomizer Spray Dryer

Article · May 2012

CITATIONS READS
12 286

3 authors, including:

Farah Saleena Taip Nuraini Abdul Aziz

Universiti Putra Malaysia Universiti Putra Malaysia
103 PUBLICATIONS   772 CITATIONS    115 PUBLICATIONS   605 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Breakwater Model Simulation Base on Mangrove Structure View project

Drying of pink guava View project

All content following this page was uploaded by Farah Saleena Taip on 23 February 2016.

The user has requested enhancement of the downloaded file.

International Journal of Engineering & Technology IJET Vol: 9 No: 10
Simulation and Control of Spray Drying Using Nozzle Atomizer
Spray Dryer
L.W. Tan, F.S. Taip *, N. Abdul Aziz
Abstract-Spray drying is a commonly used method of drying a
liquid feed through a hot gas. This study aims to obtain the
empirical model of the spray drying process of full cream milk
using a nozzle atomizer spray dryer, Lab-Plant SD 05
Laboratory Scale Spray Dryer. Inlet air temperature was
chosen as the manipulated variable and outlet air temperature
was the controlled variable. No disturbance was considered in
this process. The model was obtained from empirical model
development and it can be represented using first order plus
time delay (FOPTD). The empirical dynamic model of the
spray drying of full cream milk was simulated using
SIMULINK to evaluate the performance and robustness. The
PI and PID controllers were applied to implement the control
strategies of the process. The effects of parameter
uncertainties were investigated. From the observation, the
direct synthesis tuning method has been found as a good
controller tuning for both controllers in spray drying control
system.
Index Terms-Spray drying, PID controller, simulation, process
control
I. INTRODUCTION
Spray drying is the transformation of fluid state or pasted
material into dried particulate form by spraying the feed into a hot
drying medium where the moisture is evaporated [1][2]. Spray
drying can forms a powdered spherical, granule or agglomerated
product directly depending upon the physical and chemical
properties of the feed, the dryer design and operation [3]. Spray
dryers represent the important unit operation and are widely used
in the chemical, pharmaceutical and dairy industries where heat-
sensitive materials have to be dried [4]. In the dairy industry, this
technology is the core component for the dried milk products
production plant [5].
Dr. F.S. Taip, PhD, is working as senior lecturer at Department of
Process and Food Engineering, Faculty of Engineering, Universiti Putra
Malaysia, Malaysia. Her area of interest is in the field of Process Control
and Food Process Engineering. She is the corresponding author. (Phone:
+603-89466357; Fax: +603-86567123; email: saleena@upm.edu.my)
Dr. N. Abdul Aziz, PhD, is a senior lecturer at Department of Process
and Food Engineering, Faculty of Engineering, Universiti Putra Malaysia,
Malaysia. Her major fields of research interests are Fouling and Cleaning
and Mathematical Models for Processes (email: shekeen@eng.upm.my).
L. W. Tan received the B. Eng. in Universiti Putra Malaysia and
currently pursues her M.Sc. in Food Engineering at Universiti Putra
Malaysia, Malaysia. Her current research interests include dynamic
modeling, simulation and controlling (email: lwtan_85@yahoo.com.sg).
90710-2929 IJET-IJENS @International Journals of Engineering and Sciences IJENS
12
In the food processing industry, one of the primary objectives
of process control is to maintain the desired product quality. In
spray drying process control system, the control of moisture
content is the key to maintain the dried product’s quality.
According to [6], they found that on line measurement of dried
product moisture content is difficult due to lack of robust sensors.
Product moisture content is the most effective product parameter
and could be influenced by operating parameters. The inlet air
temperature is a major and important variable affecting the
moisture content of the dried product [7]. Increasing the inlet air
temperature will decrease the moisture content of the dried
product [7][8].
The design of a good control system rests on appropriate
understanding the dynamic behavior of the process. This requires
the development of effective dynamic models. The steps in spray
drying control system design include the selection of controlled
and manipulated variables; selecting the control strategy and the
specification of controller settings. Once the control system is
installed, the controller settings need to be tuned; therefore the
control system can perform satisfactorily [9]. The dynamic model
is never perfect because of errors in the model identification
procedure. Changes of operating conditions affect the dynamic
behavior of the feedback process and indirectly affecting the dried
product quality.
This study is focused on spray drying process using nozzle
atomizer spray dryer. The objectives of this work are to simulate
the dynamic model and propose a controller with suitable tuning
method that it satisfies both robustness and optimum performance
requirements. Nowadays the majority of industrial plants work
under control, where performance and robustness are of great
importance. This is because poor control system may results in
damages, losses and waste of significant cost. Consequently,
model uncertainty and robustness are central themes in the field of
process control [10].
II. METHODOLOGY
The dynamic model studies were focused on nozzle atomizer
spray dryer (Lab-Plant SD 05 Laboratory Scale Spray Dryer
manufactured by L.P. Technology LTD, England). The full cream
milk (UHT milk, manufactured from Dutch Lady, Malaysia) was
used as a feed material. The spray dryer was controlled manually
with constant feed flow rate. The control objective of spray dryer
control system was to keep the outlet air temperature, Tout (process
output) at the set-point by manipulating the inlet air temperature,
Tin (process input).
The data of step change in inlet air temperature which were
obtained from experiment will be used as the main source for
empirical model development. The control strategy used was
feedback control system that involved PI and PID controllers and
International Journal of Engineering & Technology IJET Vol: 9 No: 10 13

tuned by four tuning methods which were direct synthesis (M1) that corresponds to closed loop speed of response. As τc
[9], Rovira (M2) [11], Chidambaram (M3) [12] and Ziegler- decreased, Kc increased causing the closed loop to become
Nichols (M4) [11]. M1 is a model-based technique and it is unstable. If a larger τc is specified, the more sluggish response is
recommended for control system design. M2 is referred as a expected. The choice of τc is a key decision in M1. Simulations
minimum error integral tuning. M3 method is a simple method to were done to fine tune M1 to obtain the best suggested value of τc
design PI and PID controllers for stable FOPTD systems and M4 for this particular process. Based on the results, τc= θ was found to
is a classical method used as a preliminary design for designing be the most suitable value that achieved satisfactory performance
PID controllers. for both PI and PID controllers. This can be seen in Figure 1
The performances of controller settings obtained by different where the output responses were the fastest with small overshoot.
tuning methods were analyzed by using step responses of the
closed loop system and the overshoot, rise time, settling time TABLE 1: TUNING PARAMETERS FOR PI AND PID
(based on time required to reach ±5% of the steady state value) CONTROLLERS TUNED BY M1
and integral error measures (integral of the absolute value of the
error, IAE and integral of square of the error, ISE) were Closed loop PI controller PID controller
compared. The robustness of this control system was analyzed by time constant, τc Kc τI Kc τI τD
comparing the relative stability metrics: gain margin (GM) and
phase margin (PM). The effects of model uncertainties were θ 1.7000 6.8000 2.4333 7.3000 0.4658
evaluated by applying ±25% change from the process model
0.25τ 1.2593 6.8000 1.6591 7.3000 0.4658
nominal values.
0.5τ 0.7727 6.8000 0.9359 7.3000 0.4658

III. RESULTS AND DISCUSSION

3.1 Dynamic of Empirical Model
An empirical model of Lab-Plant SD 05 Laboratory Scale
Spray Dryer was developed by using the data of step change in inlet
air temperature and plotted by using cftool in MATLAB program.
The spray dryer was controlled manually with constant feed flow
rate. The predicted curve shows the first order plus time delay
(FOPTD) model. The FOPTD model parameters can be obtained by
graphic analysis of the process reaction curve. The process gain, Kp
=2 which is determined by ratio of outlet air temperature changes to
the inlet temperature change. Time delay, θ=1 minute, was
determined as the intersection of the tangent with the maximum
slope with the time axis. Time constant, τ=6.8 minutes, was
determined as time at 63.2% of output process minus θ. The values
of Kp, τ and θ are substituted into the Equation (1) which shows the
transfer function for FOPTD model. The obtained model is
applicable since the ratio  /   is 0.147, therefore the proposed
existing tuning methods can be extrapolated.
2e 1s
G p ( s)  (1)
6.8s  1

3.2 Simulation Studies

The simulation studies were done using SIMULINK from
MATLAB software. For M1 tuning method, the closed loop time
constant, τc was the single tuning parameter to be selected. Several
values of τc were suggested in the literature. As a rule of thumb,
τc=0.5τ was chosen. According to [13], τc=0.25τ was suggested
and [9] proposed a suggested value with τc=θ. Table 1 showed the
values of tuning parameters obtained from M1 using MATLAB
Command for both PI and PID controllers. It can be concluded
that the closed loop time constant, τc is inversely proportional to
controller gain, Kc. Other tuning parameters such as integral time,
τI and derivative time, τD were not affected by τc.
The responses of closed loop process tuned by M1 with Figure 1: The process responses tuned by M1 for (a) PI
different closed loop time constant, τc for both PI and PID controller and (b) PID controller
controllers were shown in Figure 1. τc is user-specified parameter

90710-2929 IJET-IJENS @International Journals of Engineering and Sciences IJENS

International Journal of Engineering & Technology IJET Vol: 9 No: 10 14

Four tuning methods were used to tune both types of are significantly the lowest. However, the controllers tuned using
controllers. As observed in Table 2, the tuning parameters values this method took the longest time to achieve the set point value. A
obtained from these tuning methods for both PI and PID suitable control strategy used in spray drying process is feedback
controllers. Table 2 showed integral time, τI was inversely control system that involved PI controller because it gave a high
proportional to controller gain, Kc. PI and PID controllers tuned degree and consistency performance compared to feedback
by M1 and M2 had smaller controller gain, Kc and larger integral control system that involved PID controller. In theory, control
time, τI corresponding to slower response. On the other hand, the performance could be improved by the use of derivative action.
controllers tuned by M3 and M4 had larger Kc and smaller τI That performance analysis showed the derivative action has small
therefore generally resulted in oscillatory response as expected. effect of performance improvement. Refer to Table 3, PID
controller just reduced around 5% overshoot, 30 seconds rise time
and 2 minutes settling time.
TABLE 2: TUNING PARAMETERS FOR PI AND PID
CONTROLLERS TUNED BY DIFFERENT TUNING
METHODS
Tuning PI controller PID controller
method Kc τI Kc τI τD
M1 1.7000 6.8000 2.4333 7.3000 0.4658
M2 1.9744 6.9923 2.8724 9.4329 0.4104
M3 2.6839 5.3084 3.4597 2.9569 0.5225
M4 3.0600 3.3300 4.0800 2.0000 0.5000

In Figure 2, the open and closed loop responses due to step

point change have been compared for this system. As expected,
the open loop response (without controller) did not reach set point
and giving 100% offset. The closed loop processes (with
controller) had desirable dynamic and steady state response
characteristics. M3 and M4 resulted in oscillatory while M1 and
M2 resulted in overdamped responses before reach steady state.
Therefore, M1 and M2 can be considered as the suitable tuning
methods to control both PI and PID controller for this spray
drying control system. For controller strategies, PI controller gave
more stable and less oscillatory response than PID controller
because PI controller had smaller Kc and larger τI than PID
controller.

3.3 Performance Analyses

The output control performances of spray drying process with
obtained FOPTD dynamics were analyzed and shown in Table 3.
The well known performance criteria such as overshoot, rise time,
settling time, IAE and ISE.
From Table 3, it is observed that the overshoot (OS) was
proportional to settling time (ts) but inversely proportional to rise
time (tr). As overshoot increases, rise time tends to decrease but
settling time, IAE and ISE increases. This situation was strongly
related to the controller gain, Kc. Controller tuned using M4
tuning method resulted in largest Kc (refer to Table 2), therefore
the output response exhibited the largest overshoot, fastest rise
time, longest settling time and produces large values of IAE and
ISE. The largest IAE represents that the largest sum of area below
and above the set point and the larger ISE indicates that poorer
performance of controller tuned using M4. M1 resulted in the
smallest Kc, which caused the sluggish output response and with
the lowest overshoot, longest rise time and fastest settling time for
both PI and PID controllers.
For the purpose of controlling the spray drying process, the Figure 2: The process response tuned by tuning methods
applicable tuning method is M1. The overshoot, settling time, IAE for (a) PI and (b) PID controller

90710-2929 IJET-IJENS @International Journals of Engineering and Sciences IJENS

 International Journal of Engineering & Technology IJET Vol: 9 No: 10 15

TABLE 3: PERFORMANCE CHARACTERISTICS FOR PI AND PID CONTROLLERS TUNED BY DIFFERENT TUNING
METHODS
Tuning PI controller PID controller
method
OS tr ts IAE ISE OS tr ts IAE ISE
M1 4.0400 3.7490 3.3950 2.1690 2.5040 3.2000 4.3880 2.8740 2.1100 2.1780
M2 9.4200 3.0550 5.1020 2.1120 2.3740 3.8000 2.4370 2.2610 1.6900 2.0670
M3 34.7600 2.2700 6.6780 2.4400 2.2340 35.0000 1.9080 8.0390 2.5150 2.0940
M4 62.2800 2.0390 9.6780 3.2380 2.4750 69.0000 1.7390 7.1640 2.7910 2.3230

3.4 Robustness Analyses
For a fair comparison of different controller settings, both
performance and robustness should be considered because these
performance criteria do not address directly another important
factor of a closed loop system-robustness. Gain and phase
margins are two well-known measures for maintaining the
robustness of the control system that indicate how close the
system is to becoming unstable. The range of GM and PM used
are (2≤GM≤5) and (30º≤PM≤75º) in PID specifications [14].
Table 4 showed the stability of PI and PID controllers checked
by calculating the GM and PM respectively.
TABLE 4: GAIN MARGIN AND PHASE MARGIN OF
CLOSED LOOP RESPONSE FOR BOTH PI AND PID
CONTROLLERS.
Tuning PI controller PID controller
method
GM PM (degree) GM PM (degree)
M1 3.1415 61.3521 2.3272 68.9515
M2 2.7100 57.1588 2.0661 63.6032
M3 1.9608 41.4557 1.5967 50.9776
M4 1.6069 27.6157 1.4259 38.3962
The larger the gain margin (GM) value shows that the
greater safety factor designed into the spray dryer control
system. The greater the safety factor resulted in the smaller the
controller gain, Kc. Therefore the controller is less sensitive to
model parameter errors. Both PI and PID controllers tuned
using M1 had the largest GM value which was resulted by the
smallest Kc and, therefore the tuned controller could withstand
greater changes in system parameters before becoming unstable.
M3 and M4 methods resulted in very low GM (<2) indicating
that those methods may not be suitable to control the process.
Phase margin (PM) indicates relative stability, the smaller
its value the more tendency for it to oscillate. M4 showed the
smallest PM value for both controllers and its response tends to
more oscillate before reaching the steady state value. Both
controllers tuned by M1 yield the largest PM, therefore the
largest phase angle can be added to the control loop before it
goes unstable.
In general, robustness can be achieved by choosing
controller setting with small Kc and large τI but it tends to result
in poor performance. It leads to tradeoff between performance
and stability of process control systems to maximize
performance while ensuring robustness. After considering both
performance and robustness, the suitable tuning method is M1.
The applicable control strategy used involves PI feedback
controller as it provides sufficient robustness and performance,
and less sensitive to errors compared to PID controller.
3.5 Process Model Mismatches
In practical situations, the dynamic model is only an
approximation of the true process dynamics. It is important to
understand how robust the control system to uncertainty of the
process dynamics. Error in the process model parameters lead to
poorer control performance and potential instability. Since the
tuning based on model, the tuning procedure should account for
the errors to provide acceptable control performance as the
process dynamics change. Therefore ±25% uncertainty in each
nominal model parameters was calculated.
Figure 3, 4 and 5 presented the closed loop responses for PI
controller after mismatch in process gain, time constant and
time delay respectively. All closed loop responses became
slower and stable when time constant, τ was increased, process
gain, Kp or time delay, θ were decreased. As expected, the
closed loop responses will become oscillatory if τ was
decreased, Kp or θ were increased.
Generally, stability problem arises for PI and PID
controllers tuned using M3 and M4 methods, especially for PID
controller tuned by M4 where the output became unstable. In
term of stability, the closed loop responses for all controllers
tuned by M1 and M2 would give better responses and
oscillation free responses. After process model mismatch, M1
would give good performances as well as the nominal process
and the suitable control strategy is PI feedback controller. The
performances of PI controller are acceptable and consistent
when process model mismatch was considered.
IV. CONCLUSION
The empirical model of nozzle atomizer spray dryer was a first
order plus time delay (FOPTD) model. PI controller tuned by
M1 is suitable due to best achievable controller performances
and highly robust when the nominal process parameters are
different than that used for controller design. The process
becomes slower if time constant increased, and process gain and
time delay decreased. On the other hand, the process will
become oscillatory when time constant decreased, and process
gain and time delay increased.

90710-2929 IJET-IJENS @International Journals of Engineering and Sciences IJENS

 International Journal of Engineering & Technology IJET Vol: 9 No: 10 16

Figure 3: Mismatch in process gain, Kp for PI controller with -25%, Figure 4: Mismatch in time constant, τ for PI controller with -
nominal and +25% respectively. 25%, nominal and +25% respectively.

90710-2929 IJET-IJENS @International Journals of Engineering and Sciences IJENS

 International Journal of Engineering & Technology IJET Vol: 9 No: 10 17

ACKNOWLEDGEMENT
We wish to thank the Department of Process and Food
Engineering, Faculty of Engineering, Universiti Putra Malaysia
(UPM) for their support.

REFERENCES
[1] Jayasundera, M., Ahhikari, B., Aldred, P., & Ghandi, A.
(2009). Surface modification of spray dried food and
emulsion powders with surface-active protein: A review.
Journal of Food Engineering, 93, 266-277.
[2] Nijdam, J. J., Guo, B., Fletcher, D. F., & Langrish, T. A. G.
(2006). Lagrangian and Eulerian models for simulating
turbulent dispension and coalescence of droplets within a
spray. Applied Mathematical Medelling 30, 1196-1211.
[3] Masters, K. (1985). Spray Drying Handbook (4th edition
ed.). London: George Godwin.
[4] Patel, K. C., & Chen, X. D. (2005). Prediction of spray-
dried product quality using two simple drying kinetics
models. Journal of Food Process Engineering, 28, 567-
594.
[5] Langrish, T. A. G. (2009). Muti-scale mathematical
modelling of spray dryers. Journal of Food Engineering,
93, 218-228.
[6] Pérez-Correa, J. R., & Zaror, C. A. (1993). Recent
advances in process control and their potential applications
to food processing Food Control, 4(4), 202-209.
[7] Chiou, D., & Langrish, T. A. G. (2008). A comparison of
cystallisation approaches in spray drying. Journal of Food
Engineering, 88, 177-185.
[8] Langrish, T. A. G. (2009). Muti-scale mathematical
modelling of spray dryers. Journal of Food Engineering,
93, 218-228.
[9] Seborg, D. E., Edgar, T. F., & Melichamp, D. A. (2004).
Process Dynamics and Control (2 ed.). New York: John
Wiley.
[10] Vijaya Selvi, J. A., Radhakrishnan, T. K., & Sundaram, S.
(2007). Performance assessment of PID and IMC tuning
methods for a mixing process with time delay. ISA
Transactions, 46, 391-397.
[11] Smith, C. A., & Corropio, A. B. (2006). Principle and
Practice of Automatic Control (3 ed.). New York: John
Wiley.
[12] Padma Sree, R., Srinivas, M. N., & Chidambaram, M.
(2004). A simple method of tuning PID controllers for
stable and unstable FOPTD systems. Computer and
Chemical Engineering, 28, 2201-2218.
[13] Tan, W., Liu, J., Chen, T., & Marquez, H. J. (2006).
Comparison of some well-known in PID tuning formulas.
Computer and Chemical Engineering, 30, 1416-1423.
[14] Äström, K. J., & Hägglund, T. (1995). PID Controllers:
Theory, Design and Tuning (2 ed.). Research Triangle
Park, NC: Instrument Society of America.
Figure 5: Mismatch in time delay, θ for PI controller with -25%,
nominal and +25% respectively.

90710-2929 IJET-IJENS @International Journals of Engineering and Sciences IJENS

View publication stats